基于空域与频域稳定特征融合的离焦虹膜识别  

Defocused iris recognition based on stable feature fusion in spatial and frequency domains

作　　者：苑玮琦[1] 刘博[1] 

机构地区：[1]沈阳工业大学视觉检测技术研究所,沈阳110870

出　　处：《仪器仪表学报》2013年第10期2300-2308,共9页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60672078);国家留学基金([2010]3006);沈阳市科技计划(F10-213-1-00)资助项目

摘　　要：针对扩大景深的定焦虹膜识别系统易引入不同程度的离焦虹膜图像,增加识别错误率的问题,提出稳定特征融合解决方案。以基于序列图像的离焦虹膜识别方法为分析和验证基础,通过对不同图像离焦程度下稳定特征在空域和频域分布规律的统计研究,确定基于单幅图像注册的稳定深点区域特征与稳定傅里叶低频相位特征的提取规则,并将二者在匹配分数级进行融合完成识别。在SUT-DI离焦虹膜数据库上的实验结果表明,与融合前基于单特征的以及基于序列图像的识别方法相比,该方法取得了更佳的识别性能,等误率最高降低7.77%,证明了该方法对于图像遭受光学离焦具有鲁棒性,能够为便捷式虹膜识别系统的开发提供技术支持。Aiming at the problems that the fixed focus iris recognition system with extended depth of field can easily introduce defocused iris images with different defocusing degrees and increase the recognition error rate, this paper proposes a stable feature fusion solution. Taking the image sequence based defocused iris recognition method as the basis for analysis and verification, the proposed method statistically studies the distribution rules of the stable features in spatial domain and frequency domain under different image defocusing degrees, determines the extraction rules of the stable dark region features and stable Fourier low frequency phase features based on single image registration, fu- ses them in the matching score level, and finally achieves the recognition. The SUT-DI defocused iris image database was used as the benchmark to carry out experiment. The experimental results show that, compared with the method based on single feature before fusion and the method based on image sequence ,the proposed method achieves better recognition performance, and the equal error rate is reduced by 7.77% at most, which proves that the proposed meth- od has robustness to the image optical defocusing, and is able to provide the technical support for the development of portable iris recognition system.

关 键 词：虹膜识别 景深 离焦虹膜图像 稳定特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]